I have a feeling that using AI-assisted coding tools (Claude Code, Cursor) gives you a much better intuition for building AI products, and that this becomes a compounding advantage for engineering teams that use AI-assisted coding tools regularly.
If you spend a lot of time using AI-assisted coding tools, all of the following starts to feel very second nature over time:
— Building up context for a specific task
— Planning out a task to various degrees of granularity
— Executing a task, and getting a sense for the level of autonomy to allow the AI
— Doing review cycles where you ask the AI to review its own work
— Identifying when the AI gets stuck, and when you need to clear the context and start over
— Tool use
Each of these tasks are things you do over and over and over again while doing AI-assisted coding, and these are the exact same types of tasks that you piece together to build an agentic AI workflow.
I have a feeling that comfort and intuition with these tasks would take much, much longer to develop if you weren’t doing AI-assisted coding regularly.
If that’s the case, what it means is that engineering teams that are using AI-assisted coding tools to build AI products would have a compounding advantage — they’re going to be faster at writing code, and also faster at developing an intuition of the capabilities of AI tools.
And for teams using a traditional software engineering approach, this would be a compounding disadvantage — they’re going to be slower at writing code and also slower at developing intuition with all of the nuances around current AI tooling.